AI Helps Astrophysicists Refine Universe's Cosmological Parameters

Researchers at the Flatiron Institute and collaborators use AI to achieve unprecedented precision in calculating the universe's cosmological parameters.

AI Enhances Understanding of Cosmological Parameters

The standard model of the universe is defined by six key numbers, known as cosmological parameters, which dictate the universe's behavior on a large scale. A team of astrophysicists, including researchers from the Flatiron Institute, has developed an AI-powered method to extract hidden information from galaxy distributions, allowing them to estimate five of these parameters with remarkable precision.

Significant Improvement Over Traditional Methods

The new AI-driven approach, dubbed Simulation-Based Inference of Galaxies (SimBIG), significantly outperforms conventional techniques. By analyzing the same galaxy data, the AI method reduced uncertainty for the parameter describing the clumpiness of matter in the universe by more than half. These results closely align with estimates derived from other cosmic observations, such as the universe's oldest light.

The researchers, led by ChangHoon Hahn from Princeton University, introduced SimBIG in a series of recent papers, including a study published on August 21 in Nature Astronomy.

A New Era of Precision in Cosmology

Generating tighter constraints on cosmological parameters is crucial for advancing our understanding of dark matter and dark energy, says study co-author Shirley Ho, a group leader at the Flatiron Institute's Center for Computational Astrophysics (CCA) in New York City. This precision will become even more vital as new cosmic surveys come online in the coming years.

"These surveys cost hundreds of millions to billions of dollars," Ho explains. "The main goal is to better understand these cosmological parameters, which are, in a sense, worth tens of millions of dollars each. Our AI method ensures that we extract the maximum amount of knowledge from these surveys, pushing the boundaries of our understanding of the universe."

Leveraging AI for Small-Scale Data Extraction

The six cosmological parameters describe various aspects of the universe, including the proportions of ordinary matter, dark matter, and dark energy, as well as the conditions following the Big Bang. Traditionally, cosmologists have calculated these parameters by analyzing the large-scale distribution of galaxies. However, this approach often overlooks valuable information on smaller scales.

"We've known for some time that there's additional information on small scales, but we didn't have an effective way to extract it," says Hahn.

The breakthrough came when Hahn proposed using AI to capture this small-scale information. The team trained their AI model using 2,000 simulated universes from the CCA-developed Quijote simulation suite. Each simulated universe was created with different values for the cosmological parameters, and the simulations were designed to mimic real galaxy survey data, including imperfections from the atmosphere and telescopes.

AI's Impact on Future Cosmic Discoveries

Once trained, the AI model analyzed data from 109,636 real galaxies, utilizing both small-scale and large-scale details to produce highly precise estimates of the cosmological parameters. The precision achieved by SimBIG was equivalent to a traditional analysis using four times as many galaxies.

This increased precision with less data is particularly valuable, says Ho, because the universe has a finite number of galaxies. SimBIG's ability to push the limits of what's possible with available data opens up exciting new possibilities in cosmology.

One of the most anticipated applications of this precision is in addressing the Hubble tension—a discrepancy between different methods of measuring the Hubble constant, which describes the rate at which the universe is expanding. By pairing data from upcoming cosmic surveys with SimBIG, researchers hope to resolve the Hubble tension or uncover new physics related to dark energy and the universe's expansion.

Collaborative Effort and Future Prospects

The study was a collaborative effort involving researchers from the Flatiron Institute, Princeton University, and several other institutions. Team members included Michael Eickenberg, Pablo Lemos, Chirag Modi, Bruno Régaldo-Saint Blancard, David Spergel, Jiamin Hou, Elena Massara, and Azadeh Moradinezhad Dizgah.

As new cosmic surveys capture more of the universe's history, the integration of AI tools like SimBIG will be crucial in revealing deeper insights into the fundamental workings of the cosmos.